Implementation and effect evaluation of dynamic difficulty adjustment based on reinforcement learning in Multiplayer Online Battle Arena games

This paper explores the implementation and impact of Dynamic Difficulty Adjustment (DDA) in Multiplayer Online Battle Arena (MOBA) games using reinforcement learning algorithms. It evaluates the effectiveness of DDA in improving player experience by dynamically adjusting game difficulty based on real-time player performance data. The study conducts empirical research on League of Legends, involving players of varying skill levels and using algorithms such as DQN, PPO, A2C, SAC, and TD3. The results indicate significant improvements in player engagement, satisfaction, and retention with the application of DDA. The paper suggests further optimization strategies for DDA and discusses its long-term effects on player behavior, highlighting the potential of reinforcement learning in game design.